vstinner
ignore `ESRCH` on processes to be killed
This patch has been rejected but I'm still using it and this pull request will serve a new discussion. From the stacktrace below you can see that it's debugger.quit() procedure we are talking about. Anything can happen during that procedure as we are killing a number of processes that depend on each other and do all sorts of stuff including quitting themselves before being killed by the debugger.
In my opinion if the caller asks the debugger to quit, the debugger should be able to do that. Not catching an exception means that debugger.quit() doesn't even finish its job and bad things can happen like left over processes and stuff like that. If special care is expected on the caller side, debugger.quit() documentation should probably be updated.
Sometimes a debugged process is already gone when python-ptrace performs
steps to terminate it. In that case we need to gracefully ignore any
ESRCH errors.
This issue was discovered when running network-testing1 and resulted in the following traceback.
Traceback (most recent call last):
File "./test-client-server", line 5, in <module>
main()
File "/home/pavlix/oss/network-testing/network_testing/client_server.py", line 55, in main
suite.run()
File "/home/pavlix/oss/network-testing/network_testing/test_suite.py", line 431, in run
testcase.run()
File "/home/pavlix/oss/network-testing/network_testing/test_suite.py", line 391, in run
scenario.run()
File "/home/pavlix/oss/network-testing/network_testing/test_suite.py", line 172, in run
debugger.quit()
File "/home/pavlix/oss/network-testing/network_testing/debug.py", line 198, in quit
super(SyscallDebugger, self).quit()
File "/usr/lib64/python3.4/site-packages/ptrace/debugger/debugger.py", line 105, in quit
process.terminate()
File "/usr/lib64/python3.4/site-packages/ptrace/debugger/process.py", line 330, in terminate
self.waitExit()
File "/usr/lib64/python3.4/site-packages/ptrace/debugger/process.py", line 351, in waitExit
self.cont(signum)
File "/usr/lib64/python3.4/site-packages/ptrace/debugger/process.py", line 717, in cont
ptrace_cont(self.pid, signum)
File "/usr/lib64/python3.4/site-packages/ptrace/binding/func.py", line 212, in ptrace_cont
ptrace(PTRACE_CONT, pid, 0, signum)
File "/usr/lib64/python3.4/site-packages/ptrace/binding/func.py", line 148, in ptrace
raise PtraceError(message, errno=errno, pid=pid)
ptrace.error.PtraceError: ptrace(cmd=7, pid=12894, 0, 133) error #3: No such process
We are in no hurry, I just wanted to make sure it isn't forgotten during the switch to github.
Your patch looks wrong: I see waitExit() in the traceback. The purpose of the function is to wait for the completion of the process, the exception should be handled here.
@haypo Just to make sure, “handled here” means handled inside the waitExit() function, i.e. waitExit() should return successfully whether the process exists or not?